Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-10-13
13:58:34 mriedem Laurene: done
14:01:11 Laurene mriedem:That's great! thanks very much!
14:01:14 melwitt mriedem: it increases contention for the cell cache so should probably revert it
14:02:11 superdan mriedem: it doesn't fix anything, but it does mean that the threads get untargeted contexts and are required to target them themselves, which is a little less clean I think
14:10:17 superdan mriedem: I'm hoping that this will get rid of the transaction context warning, and yeah I can track down the re-targeting warning in later patches
14:10:40 superdan mriedem: would you rather me separate the fix and the warnings so we don't backport the warnings?
14:10:49 superdan probably makes sense
14:14:53 openstackgerrit Dan Smith proposed openstack/nova master: Regenerate context during targeting, and sanity check some things https://review.openstack.org/511651
14:14:54 openstackgerrit Dan Smith proposed openstack/nova master: WIP: Add cell retargeting warnings https://review.openstack.org/511864
14:15:32 mriedem yeah
14:15:36 mriedem sapd_: are you around?
14:15:55 mriedem sapd_: if you're still having that problem, can you see if https://review.openstack.org/511651 fixes it for you?
14:16:40 mriedem superdan: had a couple of nits in here https://review.openstack.org/#/c/511651/4/nova/tests/unit/test_context.py
14:17:04 superdan mriedem: I looked for assertis
14:17:33 superdan ah yeah I see it
14:17:47 superdan three uses among hundreds of is not none, so I missed it
14:19:49 mriedem leaving another comment, sec
14:21:03 superdan mriedem: hurry up, I want to get to saying "wtf" to a bunch of fried_rice's patch
14:22:51 mriedem done
14:23:01 mriedem just was asking for a comment above the from_dict/to_dict dance,
14:23:13 mriedem i had to make sure that did the thing we wanted,
14:23:18 mriedem and that it somehow didn't invalidate our caching
14:23:38 bauwser mriedem: thanks for having taken time to summarize the outcome,+2d
14:23:46 mriedem bauwser: thanks
14:24:16 mriedem sdague: happy with the changes here? https://review.openstack.org/#/c/501017/
14:24:32 gibi Why does the compute manager only want to clean up an evacuated instance at startup if the virt driver still report such instance? https://github.com/openstack/nova/blob/d9212edb8fe8376c218cc40e0bee1540083b25e6/nova/compute/manager.py#L655
14:25:01 openstackgerrit Ed Leafe proposed openstack/nova master: Return Selection objects from the scheduler driver https://review.openstack.org/495854
14:25:23 gibi I could imagine situations when a compute is recovered in a way that the hypervisor does not see the instance any more but the allocations are still not cleaned up
14:25:28 sdague mriedem: sold
14:27:19 mriedem gibi: hmm, probably just legacy behavior because the RT only cared about things running on it at that time
14:27:45 mriedem the RT will pull instances based on host, and if the instance was evacuated to another host, the old recovered source RT won't find those instances and track them anymore
14:28:37 gibi but as the rt periodic task will not clean up these instances any more we have to clean up evacuated instance even if they are not visible to the hypervisor any more
14:28:50 gibi at startup explicitly
14:29:23 gibi I guess I need to create a functional test for this case as well
14:29:53 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Send Allocations to spawn https://review.openstack.org/511879
14:29:57 openstackgerrit Dan Smith proposed openstack/nova master: Regenerate context during targeting, and sanity check some things https://review.openstack.org/511651
14:29:58 openstackgerrit Dan Smith proposed openstack/nova master: WIP: Add cell retargeting warnings https://review.openstack.org/511864
14:30:03 mriedem gibi: so i assume you're thinking about moving the RP allocation cleanup outside of that for loop and add a new for loop on the 'evacuations' instances right?
14:30:15 gibi mriedem: something like that
14:30:45 gibi mriedem: the driver.destroy still need to be only run for the instances that are known by the hypervisor but the resource cleanup should run on all the evacuated instances
14:31:00 mriedem gibi: yeah probably
14:31:55 gibi mriedem: I will do this change separatly from the current fix
14:32:26 gibi mriedem: as this one will need additional functional test tweaking
14:34:50 openstackgerrit Merged openstack/nova-specs master: Spec for flavor description https://review.openstack.org/501017
14:47:02 mriedem whoa, we've been translated https://etherpad.openstack.org/p/nova-ptg-queens
14:54:43 mdbooth dansmith: We have to discuss the uuid migration thing in real time, because we're talking past each other in review :) You seem to be in violent agreement with me.
14:55:55 openstackgerrit garyk proposed openstack/nova master: Add debug information to metadata requests https://review.openstack.org/511895
14:57:18 superdan mdbooth: we may agree, but your last comment says specific things I don't agree with
14:57:29 superdan mdbooth: but sure, +1h we can
15:08:44 openstackgerrit Balazs Gibizer proposed openstack/nova master: fix cleaning up evacuated instances https://review.openstack.org/510938
15:08:45 alex_xu fried_rice: this may help you a little http://specs.openstack.org/openstack/nova-specs/specs/newton/implemented/generic-resource-pools.html#scenario-1-shared-disk-storage-used-for-vm-disk-images
15:09:09 fried_rice alex_xu Roger that, thanks.
15:09:40 alex_xu fried_rice: np
15:10:45 alex_xu fried_rice: there are some functional test for shared rp also https://review.openstack.org/#/c/498737/2
15:11:14 fried_rice cool.
15:12:38 mriedem sdague: are you happy with the rebuild + keypair spec now? https://review.openstack.org/#/c/375221/
15:28:32 sdague mriedem: yes
15:35:42 stephenfin In jaypipes' absence, who could I ask to have a look at the PCI NUMA policies spec? https://review.openstack.org/#/c/361140/
15:36:23 stephenfin superdan, perhaps? ^
15:40:12 fried_rice stephenfin Trade ya for https://review.openstack.org/510244
15:40:29 stephenfin fried_rice: Deal (y)
15:41:24 stephenfin Heh :)
15:50:47 mriedem alex_xu: you can drop the -2 on this now https://review.openstack.org/#/c/379128/
15:50:50 mriedem the blueprint is approved
16:04:20 fried_rice stephenfin Done.
16:04:21 openstackgerrit Merged openstack/nova-specs master: Reset the instance keypair while rebuilding (spec) https://review.openstack.org/375221
16:09:34 openstackgerrit Matthew Booth proposed openstack/nova-specs master: Add serial numbers for local disks https://review.openstack.org/511466
16:15:34 openstackgerrit Matt Riedemann proposed openstack/nova master: Regenerate context during targeting, and sanity check some things https://review.openstack.org/511651
16:15:49 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add cell retargeting warnings https://review.openstack.org/511864
16:19:44 superdan mriedem: the title of the commit needs to lose the "and sanity check some things" bit
16:19:56 superdan I can edit in place if you want
16:20:47 openstackgerrit Dan Smith proposed openstack/nova master: Regenerate context during targeting https://review.openstack.org/511651
16:20:52 superdan boom ^
16:23:42 openstackgerrit Eric Fried proposed openstack/nova master: Send Allocations to spawn https://review.openstack.org/511879
16:24:53 johnthetubaguy mriedem: I have one outstanding question in my head on the live-migration cinder patch, otherwise I am +2
16:26:42 openstackgerrit Eric Fried proposed openstack/nova master: Send Allocations to spawn https://review.openstack.org/511879
16:27:33 fried_rice superdan ^ with the objects outta there (and DRYing deferred to the future first patch that uses the new kwarg)
16:29:35 friedrice_injera Goofy length limits. Should be at least 255c, if not a TEXT field.
16:30:34 superdan friedrice_injera: ack, I'll look in a bit when I get back from something
16:36:41 openstackgerrit zhangyangyang proposed openstack/nova master: Fix bug of py27 job failing on testtools.matchers._impl.MismatchError https://review.openstack.org/511919
16:39:53 openstackgerrit zhangyangyang proposed openstack/nova master: Fix bug of py27 job failing on testtools.matchers._impl.MismatchError https://review.openstack.org/511919
16:49:01 stephenfin friedrice_injera: I got half way through that. Will finish bright n early Monday :) (it's nearly 6pm herE)
17:17:56 mdbooth dansmith: Ironic is a problem...
17:18:30 mdbooth dansmith: But it's also potentially a problem in other virt drivers, as they might have default serial number behaviour.
17:19:00 mdbooth I think we need to have the virt driver populate the serial number in instance device metadata.
17:19:16 mdbooth And we'd suggest that bdm.uuid is the default.
17:20:14 mdbooth Incidentally, it's weird to me that the driver creates instance.device_metadata, but I guess it makes sense.
17:20:26 mdbooth (i.e. that's the current behaviour)
17:21:57 mdbooth So ironic would query the disk's current serial number and populate it in device.serial
17:22:20 mdbooth Anyway, that's a much bigger change than I was anticipating this evening. I'll do that Monday.
17:34:04 mriedem superdan: boomshakalaka
17:34:51 mriedem johnthetubaguy: looking
17:35:42 johnthetubaguy mriedem: thanks, I think the current volume attach patch fixes the problem I found
17:35:52 johnthetubaguy mriedem: no sure what we should do about that, maybe just ignore it?
17:35:55 mriedem i think i might know what you're talking about
17:36:05 mriedem if i got tripped up on the same thing
17:38:28 mriedem superdan: some unrelated trickery here? https://review.openstack.org/#/c/506419/20/nova/compute/manager.py
17:42:23 mriedem johnthetubaguy: yeah you fell into the same trap that i did
17:42:28 mriedem finding where i pointed this out
17:48:04 johnthetubaguy mriedem: I am curious why its OK

Earlier   Later